/* * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA * or visit www.oracle.com if you need additional information or have any * questions. */ /** * @test * @bug 8233032 * @summary Tests SuperWord::co_locate_pack() involving a load pack that relies on a sandwiched and moved StoreF node. * * @run main/othervm -Xbatch -XX:+IgnoreUnrecognizedVMOptions -XX:UseAVX=1 * -XX:CompileCommand=compileonly,compiler.loopopts.superword.CoLocatePack::test * compiler.loopopts.superword.CoLocatePack */ package compiler.loopopts.superword; public class CoLocatePack { public static long lFld = 10; public static float fFld = 11.2f; public int iFld = 12; public void test() { int iArr[] = new int[200]; float fArr[] = new float[200]; /* * The IR for this loop contains the following StoreF chain after unrolling once: * StoreF 1 -> StoreF 2 -> StoreF 3 -> StoreF 4 -> StoreF 5 -> StoreF 6 * * The superword algorithm creates a pack for [ StoreF 2 and 5 ] and one for [ StoreF 3 and 6 ] * (The pack [ StoreF 1 and 4 ] is filtered out). As a result, StoreF 3 and 4 are sandwiched between * StoreF 2 and 5. SuperWord::co_locate_pack() will move both after StoreF 5 to remove any dependencies * within the pack: * StoreF 1 -> [ StoreF 2 -> StoreF 5 ] -> StoreF 3 -> StoreF 4 -> StoreF 6 * * Afterwards, StoreF 4 is moved before StoreF 3 to remove any dependency within [ StoreF 3 -> StoreF 6 ] * The resulting chain looks like this: * StoreF 1 -> [ StoreF 2 -> StoreF 5 ] -> StoreF 4 -> [ StoreF 3 -> StoreF 6 ] * * When later processing a load pack depending on StoreF 4 and 5, the first and last memory state of the load pack are * determined by using the bb indices. However, those were not updated before when moving nodes around and * bb_idx(4) < bb_idx(5) still holds even though they swapped positions in the IR. Therefore, it wrongly uses the memory * state of the first load (StoreF 5) in the pack as the last memory state. As a result, the graph walk always starts * following the input of StoreF 5 (which should actually be StoreF 4) and will move beyond a loop phi as the stop * condition is never met for a node having another memory state than the first one of the load pack. Eventually a * bb index for a node outside of the loop is read resulting in an assertion failure. * * The fix uses a different approach to find the first and last memory state of a load pack without depending on bb indices. */ for (int i = 5; i < 169; i++) { fArr[i + 1] += ((long)(fFld) | 1); // StoreF 1/4 iFld += lFld; fArr[i - 1] -= 20; // StoreF 2/5 fFld += i; fArr[i + 1] -= -117; // StoreF 3/6 int j = 10; do { } while (--j > 0); iArr[i] += 11; } } public static void main(String[] strArr) { CoLocatePack _instance = new CoLocatePack(); for (int i = 0; i < 1000; i++ ) { _instance.test(); } } }